Various computer vision and machine learning researchers has been grabbed towards Image semantic segmentation. Various applications such as autonomous driving, indoor navigation, and even virtual or augmented reality systems greatly necessitate precise and proficient segmentation mechanisms. Also, deep learning methodologies are highly demanded in nearly all field or application target correlated to computer vision, comprising semantic segmentation or scene understanding. The image segmentation is attained by series process and universal segmentation method does not prevail for low resolution image in present method, but there arises assessing difficulty for performance comparison of these segmentation methods. The research work intends at combining machine learning and deep learning approaches and thereby yielding semantically precise predictions and exhaustive segmentation along with attaining computationally effectual way. Image color segmentation is achieved by machine learning and semantic labeling through deep learning. Two algorithms are utilized in this process. First algorithm deals with detecting super pixels using Modified Support Vector Machine (MSVM) based machine learning and these super pixels segmentation is done depending on textures and colors. The second algorithm utilizes convolutional Neural Network (CNN) for training color categories and thus achieving object classification into semantic labels. It is thereby substantiated that suggested semantic image segmentation methods offers great predictive accuracy in contradiction with other segmentation approaches.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Copyright (c) 2021 Scientific Hub of Applied Research in Engineering & Information Technology